polish: TeardownError aggregation, cascade tests, README lifecycle

[lesnik512]

Summary

Closes the three non-blocking follow-ups flagged during PR1/PR2 final reviews. All small, all earned by the 2026-06-05 audit cycle, none blocking.

• fix: Generalize TeardownError aggregation across instruments. PR1's final reviewer noted that only LoggingInstrument.teardown() aggregated failures into TeardownError; the other instruments raised on first failure and skipped later cleanup steps. Now uniform:
  • OpenTelemetryInstrument.teardown() — each uninstrument() call is wrapped in its own try/except; tracer provider shutdown() is also protected; failures aggregate into a single TeardownError(errors) raised at the end. Logger restoration always runs even if an instrumentor raised.
  • FastStreamLoggingInstrument.teardown() — wraps the broker params_storage write in try/finally so super().teardown() always runs.
  • SentryInstrument.teardown() — wraps sentry_sdk.flush(timeout=2) in try/finally so sentry_sdk.init() (the SDK reset) always runs.
• test: Parametrize the OTel-insecure-endpoint cascade test across all framework configs. PR2's final reviewer noted the __post_init__ super() cascade was tested only through FastAPIConfig. The new test covers FreeConfig, LitestarConfig, FastStreamConfig, and FastAPIConfig (redundant safety net) so future inheritance refactors can't silently drop the cascade for any config.
• docs: Add a "Lifecycle constraints" section to README.md. Four constraints earned by the audit cycle that aren't obvious from the API: one bootstrapper per application instance, one OpenTelemetryInstrument per process, idempotent teardown(), and TeardownError aggregation on partial failures.

Test plan

• just test — 194 passed (187 baseline + 3 instrument-agg tests + 4 cascade-parametrize cases), 100% coverage.
• just lint-ci — clean (ruff, eof-fixer, ty).
• Each commit scoped to one logical change.

Audit cycle status

This wraps up the 2026-06-05 bug-audit-v2 work. The audit closed 26 findings across PR1 (#108), PR2 (#109), and PR3 (#110). The retrospective (planning/specs/2026-06-05-bug-audit-v2-retro.md) flagged three deferred items as non-blocking follow-ups — all three land here.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Flag	Coverage Δ
unittests	100.00% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
...bootstrap/bootstrappers/faststream_bootstrapper.py	100.00% <100.00%> (ø)
..._bootstrap/instruments/opentelemetry_instrument.py	100.00% <100.00%> (ø)
lite_bootstrap/instruments/sentry_instrument.py	100.00% <100.00%> (ø)
tests/instruments/test_opentelemetry_instrument.py	100.00% <100.00%> (ø)
tests/instruments/test_sentry_instrument.py	100.00% <100.00%> (ø)
tests/test_config.py	100.00% <100.00%> (ø)
tests/test_faststream_bootstrap.py	100.00% <100.00%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.